predict.yai error "no observations"

Question

predict.yai error "no observations"

Closed this issue a year ago · 3 comments

I am trying to use the funtion "predict.yai" on new data but I get an error saying "Error in newtargets(object, newdata, al$k, al$ann) : no observations". I am not sure what the error means or what is causing the error. I have tried to find examples for predicting on new data but I could not find any.

This is a small reproducible example based on the data MoscowMtJoe:

data("MoscowMtStJoe")
x <- MoscowMtStJoe[, c("EASTING", "NORTHING", "ELEVMEAN",
                       "SLPMEAN", "ASPMEAN", "INTMEAN", "HTMEAN", "CCMEAN")]
x[, 5] <- (1 - cos((x[, 5] - 30) * pi/180))/2
names(x)[5] = "TrASP"
y <- MoscowMtStJoe[, c(1, 9, 12, 14, 18)]

# Create yai object
mal <- yai(x = x, y = y, method = "mahalanobis")

# Create a new dataset with 30 observations based on original x but adding a bit of variance
nd = x[sample(rownames(x), nrow(x)),]+runif(nrow(x), min = -0.1, max = 0.1)
nd = nd[1:30,]
rownames(nd) = 1:nrow(nd)

# predict
predict.yai(object = mal, newdata = nd)

Answer 1 · 2024-01-29T16:13:56.000Z

What are the row names on the original data? yaImpute uses row names to identify observations. So, if the row names of the original training data are named, 1:n, the first n observations in your new data will be considered to be identical to the originals and therefore not be "new". If this is the case, I suggest you change your code so that the row names of the "new data" are formed some other way, like: rownames(x)=paste0("new",1:n) I hope this is the issue because the fix is simple! Cheers, Nick

…

On Mon, Jan 29, 2024 at 6:19 AM candelas762 ***@***.***> wrote: I am trying to use the funtion "predict.yai" on new data but I get an error saying "Error in newtargets(object, newdata, al$k, al$ann) : no observations". I am not sure what the error means or what is causing the error. I have tried to find examples for predicting on new data but I could not find any. This is a small reproducible example based on the data MoscowMtJoe: data("MoscowMtStJoe") x <- MoscowMtStJoe[, c("EASTING", "NORTHING", "ELEVMEAN", "SLPMEAN", "ASPMEAN", "INTMEAN", "HTMEAN", "CCMEAN")] x[, 5] <- (1 - cos((x[, 5] - 30) * pi/180))/2 names(x)[5] = "TrASP" y <- MoscowMtStJoe[, c(1, 9, 12, 14, 18)] # Create yai object mal <- yai(x = x, y = y, method = "mahalanobis") # Create a new dataset with 30 observations based on original x but adding a bit of variance nd = x[sample(rownames(x), nrow(x)),]+runif(nrow(x), min = -0.1, max = 0.1) nd = nd[1:30,] rownames(nd) = 1:nrow(nd) # predict predict.yai(object = mal, newdata = nd) — Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABVTTTOU5QKL374BISGHVJTYQ6VW3AVCNFSM6AAAAABCPQW576VHI2DSMVQWIX3LMV43ASLTON2WKOZSGEYDKNRQGY3TQNY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Nicholas L. Crookston Forestry Research Consultant Moscow Idaho USA

Answer 2 · 2024-01-29T16:48:15.000Z

Of course Nick nailed the issue. I change the rowname assignment in your code so that it was not duplicate with the references and it worked.

rownames(nd) = paste0("n",1:nrow(nd))

Answer 3 · 2024-01-30T08:18:35.000Z

Great! It worked and computed the predictions. Thanks!

However, I have another question. I will post it in a separate issue.